src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_avx_256_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_avx_256_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_avx_256_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  75     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  83     real *           vdwioffsetptr0;
  84     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     real *           vdwioffsetptr1;
  86     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     real *           vdwioffsetptr2;
  88     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  90     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  92     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  93     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  94     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  95     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  96     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  97     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  98     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  99     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 102     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 105     real             *charge;
 106     int              nvdwtype;
 107     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 108     int              *vdwtype;
 109     real             *vdwparam;
 110     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 111     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 112     __m128i          ewitab;
 113     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 114     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 115     real             *ewtab;
 116     __m256d          dummy_mask,cutoff_mask;
 117     __m128           tmpmask0,tmpmask1;
 118     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 119     __m256d          one     = _mm256_set1_pd(1.0);
 120     __m256d          two     = _mm256_set1_pd(2.0);
 121     x                = xx[0];
 122     f                = ff[0];
 123
 124     nri              = nlist->nri;
 125     iinr             = nlist->iinr;
 126     jindex           = nlist->jindex;
 127     jjnr             = nlist->jjnr;
 128     shiftidx         = nlist->shift;
 129     gid              = nlist->gid;
 130     shiftvec         = fr->shift_vec[0];
 131     fshift           = fr->fshift[0];
 132     facel            = _mm256_set1_pd(fr->ic->epsfac);
 133     charge           = mdatoms->chargeA;
 134     nvdwtype         = fr->ntype;
 135     vdwparam         = fr->nbfp;
 136     vdwtype          = mdatoms->typeA;
 137
 138     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 139     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 140     beta2            = _mm256_mul_pd(beta,beta);
 141     beta3            = _mm256_mul_pd(beta,beta2);
 142
 143     ewtab            = fr->ic->tabq_coul_FDV0;
 144     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 145     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 146
 147     /* Setup water-specific parameters */
 148     inr              = nlist->iinr[0];
 149     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 150     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 151     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 152     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 153
 154     jq0              = _mm256_set1_pd(charge[inr+0]);
 155     jq1              = _mm256_set1_pd(charge[inr+1]);
 156     jq2              = _mm256_set1_pd(charge[inr+2]);
 157     vdwjidx0A        = 2*vdwtype[inr+0];
 158     qq00             = _mm256_mul_pd(iq0,jq0);
 159     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 160     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 161     qq01             = _mm256_mul_pd(iq0,jq1);
 162     qq02             = _mm256_mul_pd(iq0,jq2);
 163     qq10             = _mm256_mul_pd(iq1,jq0);
 164     qq11             = _mm256_mul_pd(iq1,jq1);
 165     qq12             = _mm256_mul_pd(iq1,jq2);
 166     qq20             = _mm256_mul_pd(iq2,jq0);
 167     qq21             = _mm256_mul_pd(iq2,jq1);
 168     qq22             = _mm256_mul_pd(iq2,jq2);
 169
 170     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 171     rcutoff_scalar   = fr->ic->rcoulomb;
 172     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 173     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 174
 175     sh_vdw_invrcut6  = _mm256_set1_pd(fr->ic->sh_invrc6);
 176     rvdw             = _mm256_set1_pd(fr->ic->rvdw);
 177
 178     /* Avoid stupid compiler warnings */
 179     jnrA = jnrB = jnrC = jnrD = 0;
 180     j_coord_offsetA = 0;
 181     j_coord_offsetB = 0;
 182     j_coord_offsetC = 0;
 183     j_coord_offsetD = 0;
 184
 185     outeriter        = 0;
 186     inneriter        = 0;
 187
 188     for(iidx=0;iidx<4*DIM;iidx++)
 189     {
 190         scratch[iidx] = 0.0;
 191     }
 192
 193     /* Start outer loop over neighborlists */
 194     for(iidx=0; iidx<nri; iidx++)
 195     {
 196         /* Load shift vector for this list */
 197         i_shift_offset   = DIM*shiftidx[iidx];
 198
 199         /* Load limits for loop over neighbors */
 200         j_index_start    = jindex[iidx];
 201         j_index_end      = jindex[iidx+1];
 202
 203         /* Get outer coordinate index */
 204         inr              = iinr[iidx];
 205         i_coord_offset   = DIM*inr;
 206
 207         /* Load i particle coords and add shift vector */
 208         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 209                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 210
 211         fix0             = _mm256_setzero_pd();
 212         fiy0             = _mm256_setzero_pd();
 213         fiz0             = _mm256_setzero_pd();
 214         fix1             = _mm256_setzero_pd();
 215         fiy1             = _mm256_setzero_pd();
 216         fiz1             = _mm256_setzero_pd();
 217         fix2             = _mm256_setzero_pd();
 218         fiy2             = _mm256_setzero_pd();
 219         fiz2             = _mm256_setzero_pd();
 220
 221         /* Reset potential sums */
 222         velecsum         = _mm256_setzero_pd();
 223         vvdwsum          = _mm256_setzero_pd();
 224
 225         /* Start inner kernel loop */
 226         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 227         {
 228
 229             /* Get j neighbor index, and coordinate index */
 230             jnrA             = jjnr[jidx];
 231             jnrB             = jjnr[jidx+1];
 232             jnrC             = jjnr[jidx+2];
 233             jnrD             = jjnr[jidx+3];
 234             j_coord_offsetA  = DIM*jnrA;
 235             j_coord_offsetB  = DIM*jnrB;
 236             j_coord_offsetC  = DIM*jnrC;
 237             j_coord_offsetD  = DIM*jnrD;
 238
 239             /* load j atom coordinates */
 240             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 241                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 242                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 243
 244             /* Calculate displacement vector */
 245             dx00             = _mm256_sub_pd(ix0,jx0);
 246             dy00             = _mm256_sub_pd(iy0,jy0);
 247             dz00             = _mm256_sub_pd(iz0,jz0);
 248             dx01             = _mm256_sub_pd(ix0,jx1);
 249             dy01             = _mm256_sub_pd(iy0,jy1);
 250             dz01             = _mm256_sub_pd(iz0,jz1);
 251             dx02             = _mm256_sub_pd(ix0,jx2);
 252             dy02             = _mm256_sub_pd(iy0,jy2);
 253             dz02             = _mm256_sub_pd(iz0,jz2);
 254             dx10             = _mm256_sub_pd(ix1,jx0);
 255             dy10             = _mm256_sub_pd(iy1,jy0);
 256             dz10             = _mm256_sub_pd(iz1,jz0);
 257             dx11             = _mm256_sub_pd(ix1,jx1);
 258             dy11             = _mm256_sub_pd(iy1,jy1);
 259             dz11             = _mm256_sub_pd(iz1,jz1);
 260             dx12             = _mm256_sub_pd(ix1,jx2);
 261             dy12             = _mm256_sub_pd(iy1,jy2);
 262             dz12             = _mm256_sub_pd(iz1,jz2);
 263             dx20             = _mm256_sub_pd(ix2,jx0);
 264             dy20             = _mm256_sub_pd(iy2,jy0);
 265             dz20             = _mm256_sub_pd(iz2,jz0);
 266             dx21             = _mm256_sub_pd(ix2,jx1);
 267             dy21             = _mm256_sub_pd(iy2,jy1);
 268             dz21             = _mm256_sub_pd(iz2,jz1);
 269             dx22             = _mm256_sub_pd(ix2,jx2);
 270             dy22             = _mm256_sub_pd(iy2,jy2);
 271             dz22             = _mm256_sub_pd(iz2,jz2);
 272
 273             /* Calculate squared distance and things based on it */
 274             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 275             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 276             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 277             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 278             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 279             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 280             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 281             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 282             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 283
 284             rinv00           = avx256_invsqrt_d(rsq00);
 285             rinv01           = avx256_invsqrt_d(rsq01);
 286             rinv02           = avx256_invsqrt_d(rsq02);
 287             rinv10           = avx256_invsqrt_d(rsq10);
 288             rinv11           = avx256_invsqrt_d(rsq11);
 289             rinv12           = avx256_invsqrt_d(rsq12);
 290             rinv20           = avx256_invsqrt_d(rsq20);
 291             rinv21           = avx256_invsqrt_d(rsq21);
 292             rinv22           = avx256_invsqrt_d(rsq22);
 293
 294             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 295             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 296             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 297             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 298             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 299             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 300             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 301             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 302             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 303
 304             fjx0             = _mm256_setzero_pd();
 305             fjy0             = _mm256_setzero_pd();
 306             fjz0             = _mm256_setzero_pd();
 307             fjx1             = _mm256_setzero_pd();
 308             fjy1             = _mm256_setzero_pd();
 309             fjz1             = _mm256_setzero_pd();
 310             fjx2             = _mm256_setzero_pd();
 311             fjy2             = _mm256_setzero_pd();
 312             fjz2             = _mm256_setzero_pd();
 313
 314             /**************************
 315              * CALCULATE INTERACTIONS *
 316              **************************/
 317
 318             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 319             {
 320
 321             r00              = _mm256_mul_pd(rsq00,rinv00);
 322
 323             /* EWALD ELECTROSTATICS */
 324
 325             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 326             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 327             ewitab           = _mm256_cvttpd_epi32(ewrt);
 328             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 329             ewitab           = _mm_slli_epi32(ewitab,2);
 330             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 331             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 332             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 333             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 334             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 335             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 336             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 337             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 338             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 339
 340             /* LENNARD-JONES DISPERSION/REPULSION */
 341
 342             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 343             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 344             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 345             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 346                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 347             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 348
 349             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 350
 351             /* Update potential sum for this i atom from the interaction with this j atom. */
 352             velec            = _mm256_and_pd(velec,cutoff_mask);
 353             velecsum         = _mm256_add_pd(velecsum,velec);
 354             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 355             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 356
 357             fscal            = _mm256_add_pd(felec,fvdw);
 358
 359             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 360
 361             /* Calculate temporary vectorial force */
 362             tx               = _mm256_mul_pd(fscal,dx00);
 363             ty               = _mm256_mul_pd(fscal,dy00);
 364             tz               = _mm256_mul_pd(fscal,dz00);
 365
 366             /* Update vectorial force */
 367             fix0             = _mm256_add_pd(fix0,tx);
 368             fiy0             = _mm256_add_pd(fiy0,ty);
 369             fiz0             = _mm256_add_pd(fiz0,tz);
 370
 371             fjx0             = _mm256_add_pd(fjx0,tx);
 372             fjy0             = _mm256_add_pd(fjy0,ty);
 373             fjz0             = _mm256_add_pd(fjz0,tz);
 374
 375             }
 376
 377             /**************************
 378              * CALCULATE INTERACTIONS *
 379              **************************/
 380
 381             if (gmx_mm256_any_lt(rsq01,rcutoff2))
 382             {
 383
 384             r01              = _mm256_mul_pd(rsq01,rinv01);
 385
 386             /* EWALD ELECTROSTATICS */
 387
 388             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 389             ewrt             = _mm256_mul_pd(r01,ewtabscale);
 390             ewitab           = _mm256_cvttpd_epi32(ewrt);
 391             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 392             ewitab           = _mm_slli_epi32(ewitab,2);
 393             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 394             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 395             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 396             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 397             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 398             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 399             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 400             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_sub_pd(rinv01,sh_ewald),velec));
 401             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
 402
 403             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
 404
 405             /* Update potential sum for this i atom from the interaction with this j atom. */
 406             velec            = _mm256_and_pd(velec,cutoff_mask);
 407             velecsum         = _mm256_add_pd(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 412
 413             /* Calculate temporary vectorial force */
 414             tx               = _mm256_mul_pd(fscal,dx01);
 415             ty               = _mm256_mul_pd(fscal,dy01);
 416             tz               = _mm256_mul_pd(fscal,dz01);
 417
 418             /* Update vectorial force */
 419             fix0             = _mm256_add_pd(fix0,tx);
 420             fiy0             = _mm256_add_pd(fiy0,ty);
 421             fiz0             = _mm256_add_pd(fiz0,tz);
 422
 423             fjx1             = _mm256_add_pd(fjx1,tx);
 424             fjy1             = _mm256_add_pd(fjy1,ty);
 425             fjz1             = _mm256_add_pd(fjz1,tz);
 426
 427             }
 428
 429             /**************************
 430              * CALCULATE INTERACTIONS *
 431              **************************/
 432
 433             if (gmx_mm256_any_lt(rsq02,rcutoff2))
 434             {
 435
 436             r02              = _mm256_mul_pd(rsq02,rinv02);
 437
 438             /* EWALD ELECTROSTATICS */
 439
 440             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 441             ewrt             = _mm256_mul_pd(r02,ewtabscale);
 442             ewitab           = _mm256_cvttpd_epi32(ewrt);
 443             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 444             ewitab           = _mm_slli_epi32(ewitab,2);
 445             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 446             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 447             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 448             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 449             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 450             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 451             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 452             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_sub_pd(rinv02,sh_ewald),velec));
 453             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
 454
 455             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
 456
 457             /* Update potential sum for this i atom from the interaction with this j atom. */
 458             velec            = _mm256_and_pd(velec,cutoff_mask);
 459             velecsum         = _mm256_add_pd(velecsum,velec);
 460
 461             fscal            = felec;
 462
 463             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 464
 465             /* Calculate temporary vectorial force */
 466             tx               = _mm256_mul_pd(fscal,dx02);
 467             ty               = _mm256_mul_pd(fscal,dy02);
 468             tz               = _mm256_mul_pd(fscal,dz02);
 469
 470             /* Update vectorial force */
 471             fix0             = _mm256_add_pd(fix0,tx);
 472             fiy0             = _mm256_add_pd(fiy0,ty);
 473             fiz0             = _mm256_add_pd(fiz0,tz);
 474
 475             fjx2             = _mm256_add_pd(fjx2,tx);
 476             fjy2             = _mm256_add_pd(fjy2,ty);
 477             fjz2             = _mm256_add_pd(fjz2,tz);
 478
 479             }
 480
 481             /**************************
 482              * CALCULATE INTERACTIONS *
 483              **************************/
 484
 485             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 486             {
 487
 488             r10              = _mm256_mul_pd(rsq10,rinv10);
 489
 490             /* EWALD ELECTROSTATICS */
 491
 492             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 493             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 494             ewitab           = _mm256_cvttpd_epi32(ewrt);
 495             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 496             ewitab           = _mm_slli_epi32(ewitab,2);
 497             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 498             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 499             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 500             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 501             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 502             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 503             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 504             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_sub_pd(rinv10,sh_ewald),velec));
 505             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 506
 507             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velec            = _mm256_and_pd(velec,cutoff_mask);
 511             velecsum         = _mm256_add_pd(velecsum,velec);
 512
 513             fscal            = felec;
 514
 515             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 516
 517             /* Calculate temporary vectorial force */
 518             tx               = _mm256_mul_pd(fscal,dx10);
 519             ty               = _mm256_mul_pd(fscal,dy10);
 520             tz               = _mm256_mul_pd(fscal,dz10);
 521
 522             /* Update vectorial force */
 523             fix1             = _mm256_add_pd(fix1,tx);
 524             fiy1             = _mm256_add_pd(fiy1,ty);
 525             fiz1             = _mm256_add_pd(fiz1,tz);
 526
 527             fjx0             = _mm256_add_pd(fjx0,tx);
 528             fjy0             = _mm256_add_pd(fjy0,ty);
 529             fjz0             = _mm256_add_pd(fjz0,tz);
 530
 531             }
 532
 533             /**************************
 534              * CALCULATE INTERACTIONS *
 535              **************************/
 536
 537             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 538             {
 539
 540             r11              = _mm256_mul_pd(rsq11,rinv11);
 541
 542             /* EWALD ELECTROSTATICS */
 543
 544             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 545             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 546             ewitab           = _mm256_cvttpd_epi32(ewrt);
 547             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 548             ewitab           = _mm_slli_epi32(ewitab,2);
 549             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 550             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 551             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 552             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 553             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 554             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 555             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 556             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
 557             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 558
 559             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
 560
 561             /* Update potential sum for this i atom from the interaction with this j atom. */
 562             velec            = _mm256_and_pd(velec,cutoff_mask);
 563             velecsum         = _mm256_add_pd(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 568
 569             /* Calculate temporary vectorial force */
 570             tx               = _mm256_mul_pd(fscal,dx11);
 571             ty               = _mm256_mul_pd(fscal,dy11);
 572             tz               = _mm256_mul_pd(fscal,dz11);
 573
 574             /* Update vectorial force */
 575             fix1             = _mm256_add_pd(fix1,tx);
 576             fiy1             = _mm256_add_pd(fiy1,ty);
 577             fiz1             = _mm256_add_pd(fiz1,tz);
 578
 579             fjx1             = _mm256_add_pd(fjx1,tx);
 580             fjy1             = _mm256_add_pd(fjy1,ty);
 581             fjz1             = _mm256_add_pd(fjz1,tz);
 582
 583             }
 584
 585             /**************************
 586              * CALCULATE INTERACTIONS *
 587              **************************/
 588
 589             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 590             {
 591
 592             r12              = _mm256_mul_pd(rsq12,rinv12);
 593
 594             /* EWALD ELECTROSTATICS */
 595
 596             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 597             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 598             ewitab           = _mm256_cvttpd_epi32(ewrt);
 599             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 600             ewitab           = _mm_slli_epi32(ewitab,2);
 601             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 602             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 603             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 604             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 605             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 606             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 607             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 608             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
 609             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 610
 611             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
 612
 613             /* Update potential sum for this i atom from the interaction with this j atom. */
 614             velec            = _mm256_and_pd(velec,cutoff_mask);
 615             velecsum         = _mm256_add_pd(velecsum,velec);
 616
 617             fscal            = felec;
 618
 619             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 620
 621             /* Calculate temporary vectorial force */
 622             tx               = _mm256_mul_pd(fscal,dx12);
 623             ty               = _mm256_mul_pd(fscal,dy12);
 624             tz               = _mm256_mul_pd(fscal,dz12);
 625
 626             /* Update vectorial force */
 627             fix1             = _mm256_add_pd(fix1,tx);
 628             fiy1             = _mm256_add_pd(fiy1,ty);
 629             fiz1             = _mm256_add_pd(fiz1,tz);
 630
 631             fjx2             = _mm256_add_pd(fjx2,tx);
 632             fjy2             = _mm256_add_pd(fjy2,ty);
 633             fjz2             = _mm256_add_pd(fjz2,tz);
 634
 635             }
 636
 637             /**************************
 638              * CALCULATE INTERACTIONS *
 639              **************************/
 640
 641             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 642             {
 643
 644             r20              = _mm256_mul_pd(rsq20,rinv20);
 645
 646             /* EWALD ELECTROSTATICS */
 647
 648             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 649             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 650             ewitab           = _mm256_cvttpd_epi32(ewrt);
 651             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 652             ewitab           = _mm_slli_epi32(ewitab,2);
 653             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 654             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 655             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 656             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 657             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 658             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 659             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 660             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_sub_pd(rinv20,sh_ewald),velec));
 661             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 662
 663             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 664
 665             /* Update potential sum for this i atom from the interaction with this j atom. */
 666             velec            = _mm256_and_pd(velec,cutoff_mask);
 667             velecsum         = _mm256_add_pd(velecsum,velec);
 668
 669             fscal            = felec;
 670
 671             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 672
 673             /* Calculate temporary vectorial force */
 674             tx               = _mm256_mul_pd(fscal,dx20);
 675             ty               = _mm256_mul_pd(fscal,dy20);
 676             tz               = _mm256_mul_pd(fscal,dz20);
 677
 678             /* Update vectorial force */
 679             fix2             = _mm256_add_pd(fix2,tx);
 680             fiy2             = _mm256_add_pd(fiy2,ty);
 681             fiz2             = _mm256_add_pd(fiz2,tz);
 682
 683             fjx0             = _mm256_add_pd(fjx0,tx);
 684             fjy0             = _mm256_add_pd(fjy0,ty);
 685             fjz0             = _mm256_add_pd(fjz0,tz);
 686
 687             }
 688
 689             /**************************
 690              * CALCULATE INTERACTIONS *
 691              **************************/
 692
 693             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 694             {
 695
 696             r21              = _mm256_mul_pd(rsq21,rinv21);
 697
 698             /* EWALD ELECTROSTATICS */
 699
 700             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 701             ewrt             = _mm256_mul_pd(r21,ewtabscale);
 702             ewitab           = _mm256_cvttpd_epi32(ewrt);
 703             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 704             ewitab           = _mm_slli_epi32(ewitab,2);
 705             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 706             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 707             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 708             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 709             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 710             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 711             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 712             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
 713             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
 714
 715             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
 716
 717             /* Update potential sum for this i atom from the interaction with this j atom. */
 718             velec            = _mm256_and_pd(velec,cutoff_mask);
 719             velecsum         = _mm256_add_pd(velecsum,velec);
 720
 721             fscal            = felec;
 722
 723             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 724
 725             /* Calculate temporary vectorial force */
 726             tx               = _mm256_mul_pd(fscal,dx21);
 727             ty               = _mm256_mul_pd(fscal,dy21);
 728             tz               = _mm256_mul_pd(fscal,dz21);
 729
 730             /* Update vectorial force */
 731             fix2             = _mm256_add_pd(fix2,tx);
 732             fiy2             = _mm256_add_pd(fiy2,ty);
 733             fiz2             = _mm256_add_pd(fiz2,tz);
 734
 735             fjx1             = _mm256_add_pd(fjx1,tx);
 736             fjy1             = _mm256_add_pd(fjy1,ty);
 737             fjz1             = _mm256_add_pd(fjz1,tz);
 738
 739             }
 740
 741             /**************************
 742              * CALCULATE INTERACTIONS *
 743              **************************/
 744
 745             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 746             {
 747
 748             r22              = _mm256_mul_pd(rsq22,rinv22);
 749
 750             /* EWALD ELECTROSTATICS */
 751
 752             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 753             ewrt             = _mm256_mul_pd(r22,ewtabscale);
 754             ewitab           = _mm256_cvttpd_epi32(ewrt);
 755             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 756             ewitab           = _mm_slli_epi32(ewitab,2);
 757             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 758             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 759             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 760             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 761             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 762             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 763             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 764             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
 765             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
 766
 767             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
 768
 769             /* Update potential sum for this i atom from the interaction with this j atom. */
 770             velec            = _mm256_and_pd(velec,cutoff_mask);
 771             velecsum         = _mm256_add_pd(velecsum,velec);
 772
 773             fscal            = felec;
 774
 775             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 776
 777             /* Calculate temporary vectorial force */
 778             tx               = _mm256_mul_pd(fscal,dx22);
 779             ty               = _mm256_mul_pd(fscal,dy22);
 780             tz               = _mm256_mul_pd(fscal,dz22);
 781
 782             /* Update vectorial force */
 783             fix2             = _mm256_add_pd(fix2,tx);
 784             fiy2             = _mm256_add_pd(fiy2,ty);
 785             fiz2             = _mm256_add_pd(fiz2,tz);
 786
 787             fjx2             = _mm256_add_pd(fjx2,tx);
 788             fjy2             = _mm256_add_pd(fjy2,ty);
 789             fjz2             = _mm256_add_pd(fjz2,tz);
 790
 791             }
 792
 793             fjptrA             = f+j_coord_offsetA;
 794             fjptrB             = f+j_coord_offsetB;
 795             fjptrC             = f+j_coord_offsetC;
 796             fjptrD             = f+j_coord_offsetD;
 797
 798             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 799                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 800
 801             /* Inner loop uses 432 flops */
 802         }
 803
 804         if(jidx<j_index_end)
 805         {
 806
 807             /* Get j neighbor index, and coordinate index */
 808             jnrlistA         = jjnr[jidx];
 809             jnrlistB         = jjnr[jidx+1];
 810             jnrlistC         = jjnr[jidx+2];
 811             jnrlistD         = jjnr[jidx+3];
 812             /* Sign of each element will be negative for non-real atoms.
 813              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 814              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 815              */
 816             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 817
 818             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 819             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 820             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 821
 822             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 823             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 824             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 825             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 826             j_coord_offsetA  = DIM*jnrA;
 827             j_coord_offsetB  = DIM*jnrB;
 828             j_coord_offsetC  = DIM*jnrC;
 829             j_coord_offsetD  = DIM*jnrD;
 830
 831             /* load j atom coordinates */
 832             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 833                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 834                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 835
 836             /* Calculate displacement vector */
 837             dx00             = _mm256_sub_pd(ix0,jx0);
 838             dy00             = _mm256_sub_pd(iy0,jy0);
 839             dz00             = _mm256_sub_pd(iz0,jz0);
 840             dx01             = _mm256_sub_pd(ix0,jx1);
 841             dy01             = _mm256_sub_pd(iy0,jy1);
 842             dz01             = _mm256_sub_pd(iz0,jz1);
 843             dx02             = _mm256_sub_pd(ix0,jx2);
 844             dy02             = _mm256_sub_pd(iy0,jy2);
 845             dz02             = _mm256_sub_pd(iz0,jz2);
 846             dx10             = _mm256_sub_pd(ix1,jx0);
 847             dy10             = _mm256_sub_pd(iy1,jy0);
 848             dz10             = _mm256_sub_pd(iz1,jz0);
 849             dx11             = _mm256_sub_pd(ix1,jx1);
 850             dy11             = _mm256_sub_pd(iy1,jy1);
 851             dz11             = _mm256_sub_pd(iz1,jz1);
 852             dx12             = _mm256_sub_pd(ix1,jx2);
 853             dy12             = _mm256_sub_pd(iy1,jy2);
 854             dz12             = _mm256_sub_pd(iz1,jz2);
 855             dx20             = _mm256_sub_pd(ix2,jx0);
 856             dy20             = _mm256_sub_pd(iy2,jy0);
 857             dz20             = _mm256_sub_pd(iz2,jz0);
 858             dx21             = _mm256_sub_pd(ix2,jx1);
 859             dy21             = _mm256_sub_pd(iy2,jy1);
 860             dz21             = _mm256_sub_pd(iz2,jz1);
 861             dx22             = _mm256_sub_pd(ix2,jx2);
 862             dy22             = _mm256_sub_pd(iy2,jy2);
 863             dz22             = _mm256_sub_pd(iz2,jz2);
 864
 865             /* Calculate squared distance and things based on it */
 866             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 867             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 868             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 869             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 870             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 871             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 872             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 873             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 874             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 875
 876             rinv00           = avx256_invsqrt_d(rsq00);
 877             rinv01           = avx256_invsqrt_d(rsq01);
 878             rinv02           = avx256_invsqrt_d(rsq02);
 879             rinv10           = avx256_invsqrt_d(rsq10);
 880             rinv11           = avx256_invsqrt_d(rsq11);
 881             rinv12           = avx256_invsqrt_d(rsq12);
 882             rinv20           = avx256_invsqrt_d(rsq20);
 883             rinv21           = avx256_invsqrt_d(rsq21);
 884             rinv22           = avx256_invsqrt_d(rsq22);
 885
 886             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 887             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 888             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 889             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 890             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 891             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 892             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 893             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 894             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 895
 896             fjx0             = _mm256_setzero_pd();
 897             fjy0             = _mm256_setzero_pd();
 898             fjz0             = _mm256_setzero_pd();
 899             fjx1             = _mm256_setzero_pd();
 900             fjy1             = _mm256_setzero_pd();
 901             fjz1             = _mm256_setzero_pd();
 902             fjx2             = _mm256_setzero_pd();
 903             fjy2             = _mm256_setzero_pd();
 904             fjz2             = _mm256_setzero_pd();
 905
 906             /**************************
 907              * CALCULATE INTERACTIONS *
 908              **************************/
 909
 910             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 911             {
 912
 913             r00              = _mm256_mul_pd(rsq00,rinv00);
 914             r00              = _mm256_andnot_pd(dummy_mask,r00);
 915
 916             /* EWALD ELECTROSTATICS */
 917
 918             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 919             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 920             ewitab           = _mm256_cvttpd_epi32(ewrt);
 921             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 922             ewitab           = _mm_slli_epi32(ewitab,2);
 923             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 924             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 925             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 926             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 927             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 928             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 929             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 930             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 931             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 932
 933             /* LENNARD-JONES DISPERSION/REPULSION */
 934
 935             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 936             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 937             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 938             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 939                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 940             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 941
 942             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 943
 944             /* Update potential sum for this i atom from the interaction with this j atom. */
 945             velec            = _mm256_and_pd(velec,cutoff_mask);
 946             velec            = _mm256_andnot_pd(dummy_mask,velec);
 947             velecsum         = _mm256_add_pd(velecsum,velec);
 948             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 949             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 950             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 951
 952             fscal            = _mm256_add_pd(felec,fvdw);
 953
 954             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 955
 956             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 957
 958             /* Calculate temporary vectorial force */
 959             tx               = _mm256_mul_pd(fscal,dx00);
 960             ty               = _mm256_mul_pd(fscal,dy00);
 961             tz               = _mm256_mul_pd(fscal,dz00);
 962
 963             /* Update vectorial force */
 964             fix0             = _mm256_add_pd(fix0,tx);
 965             fiy0             = _mm256_add_pd(fiy0,ty);
 966             fiz0             = _mm256_add_pd(fiz0,tz);
 967
 968             fjx0             = _mm256_add_pd(fjx0,tx);
 969             fjy0             = _mm256_add_pd(fjy0,ty);
 970             fjz0             = _mm256_add_pd(fjz0,tz);
 971
 972             }
 973
 974             /**************************
 975              * CALCULATE INTERACTIONS *
 976              **************************/
 977
 978             if (gmx_mm256_any_lt(rsq01,rcutoff2))
 979             {
 980
 981             r01              = _mm256_mul_pd(rsq01,rinv01);
 982             r01              = _mm256_andnot_pd(dummy_mask,r01);
 983
 984             /* EWALD ELECTROSTATICS */
 985
 986             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 987             ewrt             = _mm256_mul_pd(r01,ewtabscale);
 988             ewitab           = _mm256_cvttpd_epi32(ewrt);
 989             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 990             ewitab           = _mm_slli_epi32(ewitab,2);
 991             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 992             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 993             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 994             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 995             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 996             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 997             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 998             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_sub_pd(rinv01,sh_ewald),velec));
 999             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1000
1001             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
1002
1003             /* Update potential sum for this i atom from the interaction with this j atom. */
1004             velec            = _mm256_and_pd(velec,cutoff_mask);
1005             velec            = _mm256_andnot_pd(dummy_mask,velec);
1006             velecsum         = _mm256_add_pd(velecsum,velec);
1007
1008             fscal            = felec;
1009
1010             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1011
1012             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm256_mul_pd(fscal,dx01);
1016             ty               = _mm256_mul_pd(fscal,dy01);
1017             tz               = _mm256_mul_pd(fscal,dz01);
1018
1019             /* Update vectorial force */
1020             fix0             = _mm256_add_pd(fix0,tx);
1021             fiy0             = _mm256_add_pd(fiy0,ty);
1022             fiz0             = _mm256_add_pd(fiz0,tz);
1023
1024             fjx1             = _mm256_add_pd(fjx1,tx);
1025             fjy1             = _mm256_add_pd(fjy1,ty);
1026             fjz1             = _mm256_add_pd(fjz1,tz);
1027
1028             }
1029
1030             /**************************
1031              * CALCULATE INTERACTIONS *
1032              **************************/
1033
1034             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1035             {
1036
1037             r02              = _mm256_mul_pd(rsq02,rinv02);
1038             r02              = _mm256_andnot_pd(dummy_mask,r02);
1039
1040             /* EWALD ELECTROSTATICS */
1041
1042             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1043             ewrt             = _mm256_mul_pd(r02,ewtabscale);
1044             ewitab           = _mm256_cvttpd_epi32(ewrt);
1045             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1046             ewitab           = _mm_slli_epi32(ewitab,2);
1047             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1048             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1049             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1050             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1051             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1052             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1053             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1054             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_sub_pd(rinv02,sh_ewald),velec));
1055             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1056
1057             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
1058
1059             /* Update potential sum for this i atom from the interaction with this j atom. */
1060             velec            = _mm256_and_pd(velec,cutoff_mask);
1061             velec            = _mm256_andnot_pd(dummy_mask,velec);
1062             velecsum         = _mm256_add_pd(velecsum,velec);
1063
1064             fscal            = felec;
1065
1066             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1067
1068             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1069
1070             /* Calculate temporary vectorial force */
1071             tx               = _mm256_mul_pd(fscal,dx02);
1072             ty               = _mm256_mul_pd(fscal,dy02);
1073             tz               = _mm256_mul_pd(fscal,dz02);
1074
1075             /* Update vectorial force */
1076             fix0             = _mm256_add_pd(fix0,tx);
1077             fiy0             = _mm256_add_pd(fiy0,ty);
1078             fiz0             = _mm256_add_pd(fiz0,tz);
1079
1080             fjx2             = _mm256_add_pd(fjx2,tx);
1081             fjy2             = _mm256_add_pd(fjy2,ty);
1082             fjz2             = _mm256_add_pd(fjz2,tz);
1083
1084             }
1085
1086             /**************************
1087              * CALCULATE INTERACTIONS *
1088              **************************/
1089
1090             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1091             {
1092
1093             r10              = _mm256_mul_pd(rsq10,rinv10);
1094             r10              = _mm256_andnot_pd(dummy_mask,r10);
1095
1096             /* EWALD ELECTROSTATICS */
1097
1098             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1099             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1100             ewitab           = _mm256_cvttpd_epi32(ewrt);
1101             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1102             ewitab           = _mm_slli_epi32(ewitab,2);
1103             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1104             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1105             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1106             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1107             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1108             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1109             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1110             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_sub_pd(rinv10,sh_ewald),velec));
1111             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1112
1113             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1114
1115             /* Update potential sum for this i atom from the interaction with this j atom. */
1116             velec            = _mm256_and_pd(velec,cutoff_mask);
1117             velec            = _mm256_andnot_pd(dummy_mask,velec);
1118             velecsum         = _mm256_add_pd(velecsum,velec);
1119
1120             fscal            = felec;
1121
1122             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1123
1124             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1125
1126             /* Calculate temporary vectorial force */
1127             tx               = _mm256_mul_pd(fscal,dx10);
1128             ty               = _mm256_mul_pd(fscal,dy10);
1129             tz               = _mm256_mul_pd(fscal,dz10);
1130
1131             /* Update vectorial force */
1132             fix1             = _mm256_add_pd(fix1,tx);
1133             fiy1             = _mm256_add_pd(fiy1,ty);
1134             fiz1             = _mm256_add_pd(fiz1,tz);
1135
1136             fjx0             = _mm256_add_pd(fjx0,tx);
1137             fjy0             = _mm256_add_pd(fjy0,ty);
1138             fjz0             = _mm256_add_pd(fjz0,tz);
1139
1140             }
1141
1142             /**************************
1143              * CALCULATE INTERACTIONS *
1144              **************************/
1145
1146             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1147             {
1148
1149             r11              = _mm256_mul_pd(rsq11,rinv11);
1150             r11              = _mm256_andnot_pd(dummy_mask,r11);
1151
1152             /* EWALD ELECTROSTATICS */
1153
1154             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1155             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1156             ewitab           = _mm256_cvttpd_epi32(ewrt);
1157             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1158             ewitab           = _mm_slli_epi32(ewitab,2);
1159             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1160             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1161             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1162             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1163             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1164             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1165             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1166             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
1167             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1168
1169             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1170
1171             /* Update potential sum for this i atom from the interaction with this j atom. */
1172             velec            = _mm256_and_pd(velec,cutoff_mask);
1173             velec            = _mm256_andnot_pd(dummy_mask,velec);
1174             velecsum         = _mm256_add_pd(velecsum,velec);
1175
1176             fscal            = felec;
1177
1178             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1179
1180             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1181
1182             /* Calculate temporary vectorial force */
1183             tx               = _mm256_mul_pd(fscal,dx11);
1184             ty               = _mm256_mul_pd(fscal,dy11);
1185             tz               = _mm256_mul_pd(fscal,dz11);
1186
1187             /* Update vectorial force */
1188             fix1             = _mm256_add_pd(fix1,tx);
1189             fiy1             = _mm256_add_pd(fiy1,ty);
1190             fiz1             = _mm256_add_pd(fiz1,tz);
1191
1192             fjx1             = _mm256_add_pd(fjx1,tx);
1193             fjy1             = _mm256_add_pd(fjy1,ty);
1194             fjz1             = _mm256_add_pd(fjz1,tz);
1195
1196             }
1197
1198             /**************************
1199              * CALCULATE INTERACTIONS *
1200              **************************/
1201
1202             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1203             {
1204
1205             r12              = _mm256_mul_pd(rsq12,rinv12);
1206             r12              = _mm256_andnot_pd(dummy_mask,r12);
1207
1208             /* EWALD ELECTROSTATICS */
1209
1210             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1211             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1212             ewitab           = _mm256_cvttpd_epi32(ewrt);
1213             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1214             ewitab           = _mm_slli_epi32(ewitab,2);
1215             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1216             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1217             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1218             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1219             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1220             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1221             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1222             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
1223             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1224
1225             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
1226
1227             /* Update potential sum for this i atom from the interaction with this j atom. */
1228             velec            = _mm256_and_pd(velec,cutoff_mask);
1229             velec            = _mm256_andnot_pd(dummy_mask,velec);
1230             velecsum         = _mm256_add_pd(velecsum,velec);
1231
1232             fscal            = felec;
1233
1234             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1235
1236             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1237
1238             /* Calculate temporary vectorial force */
1239             tx               = _mm256_mul_pd(fscal,dx12);
1240             ty               = _mm256_mul_pd(fscal,dy12);
1241             tz               = _mm256_mul_pd(fscal,dz12);
1242
1243             /* Update vectorial force */
1244             fix1             = _mm256_add_pd(fix1,tx);
1245             fiy1             = _mm256_add_pd(fiy1,ty);
1246             fiz1             = _mm256_add_pd(fiz1,tz);
1247
1248             fjx2             = _mm256_add_pd(fjx2,tx);
1249             fjy2             = _mm256_add_pd(fjy2,ty);
1250             fjz2             = _mm256_add_pd(fjz2,tz);
1251
1252             }
1253
1254             /**************************
1255              * CALCULATE INTERACTIONS *
1256              **************************/
1257
1258             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1259             {
1260
1261             r20              = _mm256_mul_pd(rsq20,rinv20);
1262             r20              = _mm256_andnot_pd(dummy_mask,r20);
1263
1264             /* EWALD ELECTROSTATICS */
1265
1266             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1267             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1268             ewitab           = _mm256_cvttpd_epi32(ewrt);
1269             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1270             ewitab           = _mm_slli_epi32(ewitab,2);
1271             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1272             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1273             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1274             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1275             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1276             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1277             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1278             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_sub_pd(rinv20,sh_ewald),velec));
1279             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1280
1281             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1282
1283             /* Update potential sum for this i atom from the interaction with this j atom. */
1284             velec            = _mm256_and_pd(velec,cutoff_mask);
1285             velec            = _mm256_andnot_pd(dummy_mask,velec);
1286             velecsum         = _mm256_add_pd(velecsum,velec);
1287
1288             fscal            = felec;
1289
1290             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1291
1292             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1293
1294             /* Calculate temporary vectorial force */
1295             tx               = _mm256_mul_pd(fscal,dx20);
1296             ty               = _mm256_mul_pd(fscal,dy20);
1297             tz               = _mm256_mul_pd(fscal,dz20);
1298
1299             /* Update vectorial force */
1300             fix2             = _mm256_add_pd(fix2,tx);
1301             fiy2             = _mm256_add_pd(fiy2,ty);
1302             fiz2             = _mm256_add_pd(fiz2,tz);
1303
1304             fjx0             = _mm256_add_pd(fjx0,tx);
1305             fjy0             = _mm256_add_pd(fjy0,ty);
1306             fjz0             = _mm256_add_pd(fjz0,tz);
1307
1308             }
1309
1310             /**************************
1311              * CALCULATE INTERACTIONS *
1312              **************************/
1313
1314             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1315             {
1316
1317             r21              = _mm256_mul_pd(rsq21,rinv21);
1318             r21              = _mm256_andnot_pd(dummy_mask,r21);
1319
1320             /* EWALD ELECTROSTATICS */
1321
1322             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1323             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1324             ewitab           = _mm256_cvttpd_epi32(ewrt);
1325             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1326             ewitab           = _mm_slli_epi32(ewitab,2);
1327             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1328             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1329             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1330             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1331             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1332             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1333             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1334             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
1335             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1336
1337             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
1338
1339             /* Update potential sum for this i atom from the interaction with this j atom. */
1340             velec            = _mm256_and_pd(velec,cutoff_mask);
1341             velec            = _mm256_andnot_pd(dummy_mask,velec);
1342             velecsum         = _mm256_add_pd(velecsum,velec);
1343
1344             fscal            = felec;
1345
1346             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1347
1348             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1349
1350             /* Calculate temporary vectorial force */
1351             tx               = _mm256_mul_pd(fscal,dx21);
1352             ty               = _mm256_mul_pd(fscal,dy21);
1353             tz               = _mm256_mul_pd(fscal,dz21);
1354
1355             /* Update vectorial force */
1356             fix2             = _mm256_add_pd(fix2,tx);
1357             fiy2             = _mm256_add_pd(fiy2,ty);
1358             fiz2             = _mm256_add_pd(fiz2,tz);
1359
1360             fjx1             = _mm256_add_pd(fjx1,tx);
1361             fjy1             = _mm256_add_pd(fjy1,ty);
1362             fjz1             = _mm256_add_pd(fjz1,tz);
1363
1364             }
1365
1366             /**************************
1367              * CALCULATE INTERACTIONS *
1368              **************************/
1369
1370             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1371             {
1372
1373             r22              = _mm256_mul_pd(rsq22,rinv22);
1374             r22              = _mm256_andnot_pd(dummy_mask,r22);
1375
1376             /* EWALD ELECTROSTATICS */
1377
1378             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1379             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1380             ewitab           = _mm256_cvttpd_epi32(ewrt);
1381             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1382             ewitab           = _mm_slli_epi32(ewitab,2);
1383             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1384             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1385             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1386             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1387             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1388             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1389             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1390             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
1391             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1392
1393             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
1394
1395             /* Update potential sum for this i atom from the interaction with this j atom. */
1396             velec            = _mm256_and_pd(velec,cutoff_mask);
1397             velec            = _mm256_andnot_pd(dummy_mask,velec);
1398             velecsum         = _mm256_add_pd(velecsum,velec);
1399
1400             fscal            = felec;
1401
1402             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1403
1404             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1405
1406             /* Calculate temporary vectorial force */
1407             tx               = _mm256_mul_pd(fscal,dx22);
1408             ty               = _mm256_mul_pd(fscal,dy22);
1409             tz               = _mm256_mul_pd(fscal,dz22);
1410
1411             /* Update vectorial force */
1412             fix2             = _mm256_add_pd(fix2,tx);
1413             fiy2             = _mm256_add_pd(fiy2,ty);
1414             fiz2             = _mm256_add_pd(fiz2,tz);
1415
1416             fjx2             = _mm256_add_pd(fjx2,tx);
1417             fjy2             = _mm256_add_pd(fjy2,ty);
1418             fjz2             = _mm256_add_pd(fjz2,tz);
1419
1420             }
1421
1422             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1423             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1424             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1425             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1426
1427             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1428                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1429
1430             /* Inner loop uses 441 flops */
1431         }
1432
1433         /* End of innermost loop */
1434
1435         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1436                                                  f+i_coord_offset,fshift+i_shift_offset);
1437
1438         ggid                        = gid[iidx];
1439         /* Update potential energies */
1440         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1441         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1442
1443         /* Increment number of inner iterations */
1444         inneriter                  += j_index_end - j_index_start;
1445
1446         /* Outer loop uses 20 flops */
1447     }
1448
1449     /* Increment number of outer iterations */
1450     outeriter        += nri;
1451
1452     /* Update outer/inner flops */
1453
1454     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*441);
1455 }
1456 /*
1457  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_avx_256_double
1458  * Electrostatics interaction: Ewald
1459  * VdW interaction:            LennardJones
1460  * Geometry:                   Water3-Water3
1461  * Calculate force/pot:        Force
1462  */
1463 void
1464 nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_avx_256_double
1465                     (t_nblist                    * gmx_restrict       nlist,
1466                      rvec                        * gmx_restrict          xx,
1467                      rvec                        * gmx_restrict          ff,
1468                      struct t_forcerec           * gmx_restrict          fr,
1469                      t_mdatoms                   * gmx_restrict     mdatoms,
1470                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1471                      t_nrnb                      * gmx_restrict        nrnb)
1472 {
1473     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1474      * just 0 for non-waters.
1475      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1476      * jnr indices corresponding to data put in the four positions in the SIMD register.
1477      */
1478     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1479     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1480     int              jnrA,jnrB,jnrC,jnrD;
1481     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1482     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1483     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1484     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1485     real             rcutoff_scalar;
1486     real             *shiftvec,*fshift,*x,*f;
1487     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1488     real             scratch[4*DIM];
1489     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1490     real *           vdwioffsetptr0;
1491     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1492     real *           vdwioffsetptr1;
1493     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1494     real *           vdwioffsetptr2;
1495     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1496     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1497     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1498     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1499     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1500     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1501     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1502     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1503     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1504     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1505     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1506     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1507     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1508     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1509     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1510     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1511     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1512     real             *charge;
1513     int              nvdwtype;
1514     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1515     int              *vdwtype;
1516     real             *vdwparam;
1517     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1518     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1519     __m128i          ewitab;
1520     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1521     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1522     real             *ewtab;
1523     __m256d          dummy_mask,cutoff_mask;
1524     __m128           tmpmask0,tmpmask1;
1525     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1526     __m256d          one     = _mm256_set1_pd(1.0);
1527     __m256d          two     = _mm256_set1_pd(2.0);
1528     x                = xx[0];
1529     f                = ff[0];
1530
1531     nri              = nlist->nri;
1532     iinr             = nlist->iinr;
1533     jindex           = nlist->jindex;
1534     jjnr             = nlist->jjnr;
1535     shiftidx         = nlist->shift;
1536     gid              = nlist->gid;
1537     shiftvec         = fr->shift_vec[0];
1538     fshift           = fr->fshift[0];
1539     facel            = _mm256_set1_pd(fr->ic->epsfac);
1540     charge           = mdatoms->chargeA;
1541     nvdwtype         = fr->ntype;
1542     vdwparam         = fr->nbfp;
1543     vdwtype          = mdatoms->typeA;
1544
1545     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
1546     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1547     beta2            = _mm256_mul_pd(beta,beta);
1548     beta3            = _mm256_mul_pd(beta,beta2);
1549
1550     ewtab            = fr->ic->tabq_coul_F;
1551     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
1552     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1553
1554     /* Setup water-specific parameters */
1555     inr              = nlist->iinr[0];
1556     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1557     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1558     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1559     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1560
1561     jq0              = _mm256_set1_pd(charge[inr+0]);
1562     jq1              = _mm256_set1_pd(charge[inr+1]);
1563     jq2              = _mm256_set1_pd(charge[inr+2]);
1564     vdwjidx0A        = 2*vdwtype[inr+0];
1565     qq00             = _mm256_mul_pd(iq0,jq0);
1566     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1567     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1568     qq01             = _mm256_mul_pd(iq0,jq1);
1569     qq02             = _mm256_mul_pd(iq0,jq2);
1570     qq10             = _mm256_mul_pd(iq1,jq0);
1571     qq11             = _mm256_mul_pd(iq1,jq1);
1572     qq12             = _mm256_mul_pd(iq1,jq2);
1573     qq20             = _mm256_mul_pd(iq2,jq0);
1574     qq21             = _mm256_mul_pd(iq2,jq1);
1575     qq22             = _mm256_mul_pd(iq2,jq2);
1576
1577     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1578     rcutoff_scalar   = fr->ic->rcoulomb;
1579     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
1580     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
1581
1582     sh_vdw_invrcut6  = _mm256_set1_pd(fr->ic->sh_invrc6);
1583     rvdw             = _mm256_set1_pd(fr->ic->rvdw);
1584
1585     /* Avoid stupid compiler warnings */
1586     jnrA = jnrB = jnrC = jnrD = 0;
1587     j_coord_offsetA = 0;
1588     j_coord_offsetB = 0;
1589     j_coord_offsetC = 0;
1590     j_coord_offsetD = 0;
1591
1592     outeriter        = 0;
1593     inneriter        = 0;
1594
1595     for(iidx=0;iidx<4*DIM;iidx++)
1596     {
1597         scratch[iidx] = 0.0;
1598     }
1599
1600     /* Start outer loop over neighborlists */
1601     for(iidx=0; iidx<nri; iidx++)
1602     {
1603         /* Load shift vector for this list */
1604         i_shift_offset   = DIM*shiftidx[iidx];
1605
1606         /* Load limits for loop over neighbors */
1607         j_index_start    = jindex[iidx];
1608         j_index_end      = jindex[iidx+1];
1609
1610         /* Get outer coordinate index */
1611         inr              = iinr[iidx];
1612         i_coord_offset   = DIM*inr;
1613
1614         /* Load i particle coords and add shift vector */
1615         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1616                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1617
1618         fix0             = _mm256_setzero_pd();
1619         fiy0             = _mm256_setzero_pd();
1620         fiz0             = _mm256_setzero_pd();
1621         fix1             = _mm256_setzero_pd();
1622         fiy1             = _mm256_setzero_pd();
1623         fiz1             = _mm256_setzero_pd();
1624         fix2             = _mm256_setzero_pd();
1625         fiy2             = _mm256_setzero_pd();
1626         fiz2             = _mm256_setzero_pd();
1627
1628         /* Start inner kernel loop */
1629         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1630         {
1631
1632             /* Get j neighbor index, and coordinate index */
1633             jnrA             = jjnr[jidx];
1634             jnrB             = jjnr[jidx+1];
1635             jnrC             = jjnr[jidx+2];
1636             jnrD             = jjnr[jidx+3];
1637             j_coord_offsetA  = DIM*jnrA;
1638             j_coord_offsetB  = DIM*jnrB;
1639             j_coord_offsetC  = DIM*jnrC;
1640             j_coord_offsetD  = DIM*jnrD;
1641
1642             /* load j atom coordinates */
1643             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1644                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1645                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1646
1647             /* Calculate displacement vector */
1648             dx00             = _mm256_sub_pd(ix0,jx0);
1649             dy00             = _mm256_sub_pd(iy0,jy0);
1650             dz00             = _mm256_sub_pd(iz0,jz0);
1651             dx01             = _mm256_sub_pd(ix0,jx1);
1652             dy01             = _mm256_sub_pd(iy0,jy1);
1653             dz01             = _mm256_sub_pd(iz0,jz1);
1654             dx02             = _mm256_sub_pd(ix0,jx2);
1655             dy02             = _mm256_sub_pd(iy0,jy2);
1656             dz02             = _mm256_sub_pd(iz0,jz2);
1657             dx10             = _mm256_sub_pd(ix1,jx0);
1658             dy10             = _mm256_sub_pd(iy1,jy0);
1659             dz10             = _mm256_sub_pd(iz1,jz0);
1660             dx11             = _mm256_sub_pd(ix1,jx1);
1661             dy11             = _mm256_sub_pd(iy1,jy1);
1662             dz11             = _mm256_sub_pd(iz1,jz1);
1663             dx12             = _mm256_sub_pd(ix1,jx2);
1664             dy12             = _mm256_sub_pd(iy1,jy2);
1665             dz12             = _mm256_sub_pd(iz1,jz2);
1666             dx20             = _mm256_sub_pd(ix2,jx0);
1667             dy20             = _mm256_sub_pd(iy2,jy0);
1668             dz20             = _mm256_sub_pd(iz2,jz0);
1669             dx21             = _mm256_sub_pd(ix2,jx1);
1670             dy21             = _mm256_sub_pd(iy2,jy1);
1671             dz21             = _mm256_sub_pd(iz2,jz1);
1672             dx22             = _mm256_sub_pd(ix2,jx2);
1673             dy22             = _mm256_sub_pd(iy2,jy2);
1674             dz22             = _mm256_sub_pd(iz2,jz2);
1675
1676             /* Calculate squared distance and things based on it */
1677             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1678             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1679             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1680             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1681             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1682             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1683             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1684             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1685             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1686
1687             rinv00           = avx256_invsqrt_d(rsq00);
1688             rinv01           = avx256_invsqrt_d(rsq01);
1689             rinv02           = avx256_invsqrt_d(rsq02);
1690             rinv10           = avx256_invsqrt_d(rsq10);
1691             rinv11           = avx256_invsqrt_d(rsq11);
1692             rinv12           = avx256_invsqrt_d(rsq12);
1693             rinv20           = avx256_invsqrt_d(rsq20);
1694             rinv21           = avx256_invsqrt_d(rsq21);
1695             rinv22           = avx256_invsqrt_d(rsq22);
1696
1697             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1698             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1699             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1700             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1701             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1702             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1703             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1704             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1705             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1706
1707             fjx0             = _mm256_setzero_pd();
1708             fjy0             = _mm256_setzero_pd();
1709             fjz0             = _mm256_setzero_pd();
1710             fjx1             = _mm256_setzero_pd();
1711             fjy1             = _mm256_setzero_pd();
1712             fjz1             = _mm256_setzero_pd();
1713             fjx2             = _mm256_setzero_pd();
1714             fjy2             = _mm256_setzero_pd();
1715             fjz2             = _mm256_setzero_pd();
1716
1717             /**************************
1718              * CALCULATE INTERACTIONS *
1719              **************************/
1720
1721             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1722             {
1723
1724             r00              = _mm256_mul_pd(rsq00,rinv00);
1725
1726             /* EWALD ELECTROSTATICS */
1727
1728             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1729             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1730             ewitab           = _mm256_cvttpd_epi32(ewrt);
1731             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1732             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1733                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1734                                             &ewtabF,&ewtabFn);
1735             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1736             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1737
1738             /* LENNARD-JONES DISPERSION/REPULSION */
1739
1740             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1741             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1742
1743             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1744
1745             fscal            = _mm256_add_pd(felec,fvdw);
1746
1747             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1748
1749             /* Calculate temporary vectorial force */
1750             tx               = _mm256_mul_pd(fscal,dx00);
1751             ty               = _mm256_mul_pd(fscal,dy00);
1752             tz               = _mm256_mul_pd(fscal,dz00);
1753
1754             /* Update vectorial force */
1755             fix0             = _mm256_add_pd(fix0,tx);
1756             fiy0             = _mm256_add_pd(fiy0,ty);
1757             fiz0             = _mm256_add_pd(fiz0,tz);
1758
1759             fjx0             = _mm256_add_pd(fjx0,tx);
1760             fjy0             = _mm256_add_pd(fjy0,ty);
1761             fjz0             = _mm256_add_pd(fjz0,tz);
1762
1763             }
1764
1765             /**************************
1766              * CALCULATE INTERACTIONS *
1767              **************************/
1768
1769             if (gmx_mm256_any_lt(rsq01,rcutoff2))
1770             {
1771
1772             r01              = _mm256_mul_pd(rsq01,rinv01);
1773
1774             /* EWALD ELECTROSTATICS */
1775
1776             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1777             ewrt             = _mm256_mul_pd(r01,ewtabscale);
1778             ewitab           = _mm256_cvttpd_epi32(ewrt);
1779             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1780             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1781                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1782                                             &ewtabF,&ewtabFn);
1783             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1784             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1785
1786             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
1787
1788             fscal            = felec;
1789
1790             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1791
1792             /* Calculate temporary vectorial force */
1793             tx               = _mm256_mul_pd(fscal,dx01);
1794             ty               = _mm256_mul_pd(fscal,dy01);
1795             tz               = _mm256_mul_pd(fscal,dz01);
1796
1797             /* Update vectorial force */
1798             fix0             = _mm256_add_pd(fix0,tx);
1799             fiy0             = _mm256_add_pd(fiy0,ty);
1800             fiz0             = _mm256_add_pd(fiz0,tz);
1801
1802             fjx1             = _mm256_add_pd(fjx1,tx);
1803             fjy1             = _mm256_add_pd(fjy1,ty);
1804             fjz1             = _mm256_add_pd(fjz1,tz);
1805
1806             }
1807
1808             /**************************
1809              * CALCULATE INTERACTIONS *
1810              **************************/
1811
1812             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1813             {
1814
1815             r02              = _mm256_mul_pd(rsq02,rinv02);
1816
1817             /* EWALD ELECTROSTATICS */
1818
1819             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1820             ewrt             = _mm256_mul_pd(r02,ewtabscale);
1821             ewitab           = _mm256_cvttpd_epi32(ewrt);
1822             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1823             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1824                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1825                                             &ewtabF,&ewtabFn);
1826             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1827             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1828
1829             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
1830
1831             fscal            = felec;
1832
1833             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1834
1835             /* Calculate temporary vectorial force */
1836             tx               = _mm256_mul_pd(fscal,dx02);
1837             ty               = _mm256_mul_pd(fscal,dy02);
1838             tz               = _mm256_mul_pd(fscal,dz02);
1839
1840             /* Update vectorial force */
1841             fix0             = _mm256_add_pd(fix0,tx);
1842             fiy0             = _mm256_add_pd(fiy0,ty);
1843             fiz0             = _mm256_add_pd(fiz0,tz);
1844
1845             fjx2             = _mm256_add_pd(fjx2,tx);
1846             fjy2             = _mm256_add_pd(fjy2,ty);
1847             fjz2             = _mm256_add_pd(fjz2,tz);
1848
1849             }
1850
1851             /**************************
1852              * CALCULATE INTERACTIONS *
1853              **************************/
1854
1855             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1856             {
1857
1858             r10              = _mm256_mul_pd(rsq10,rinv10);
1859
1860             /* EWALD ELECTROSTATICS */
1861
1862             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1863             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1864             ewitab           = _mm256_cvttpd_epi32(ewrt);
1865             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1866             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1867                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1868                                             &ewtabF,&ewtabFn);
1869             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1870             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1871
1872             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1873
1874             fscal            = felec;
1875
1876             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1877
1878             /* Calculate temporary vectorial force */
1879             tx               = _mm256_mul_pd(fscal,dx10);
1880             ty               = _mm256_mul_pd(fscal,dy10);
1881             tz               = _mm256_mul_pd(fscal,dz10);
1882
1883             /* Update vectorial force */
1884             fix1             = _mm256_add_pd(fix1,tx);
1885             fiy1             = _mm256_add_pd(fiy1,ty);
1886             fiz1             = _mm256_add_pd(fiz1,tz);
1887
1888             fjx0             = _mm256_add_pd(fjx0,tx);
1889             fjy0             = _mm256_add_pd(fjy0,ty);
1890             fjz0             = _mm256_add_pd(fjz0,tz);
1891
1892             }
1893
1894             /**************************
1895              * CALCULATE INTERACTIONS *
1896              **************************/
1897
1898             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1899             {
1900
1901             r11              = _mm256_mul_pd(rsq11,rinv11);
1902
1903             /* EWALD ELECTROSTATICS */
1904
1905             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1906             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1907             ewitab           = _mm256_cvttpd_epi32(ewrt);
1908             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1909             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1910                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1911                                             &ewtabF,&ewtabFn);
1912             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1913             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1914
1915             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1916
1917             fscal            = felec;
1918
1919             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1920
1921             /* Calculate temporary vectorial force */
1922             tx               = _mm256_mul_pd(fscal,dx11);
1923             ty               = _mm256_mul_pd(fscal,dy11);
1924             tz               = _mm256_mul_pd(fscal,dz11);
1925
1926             /* Update vectorial force */
1927             fix1             = _mm256_add_pd(fix1,tx);
1928             fiy1             = _mm256_add_pd(fiy1,ty);
1929             fiz1             = _mm256_add_pd(fiz1,tz);
1930
1931             fjx1             = _mm256_add_pd(fjx1,tx);
1932             fjy1             = _mm256_add_pd(fjy1,ty);
1933             fjz1             = _mm256_add_pd(fjz1,tz);
1934
1935             }
1936
1937             /**************************
1938              * CALCULATE INTERACTIONS *
1939              **************************/
1940
1941             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1942             {
1943
1944             r12              = _mm256_mul_pd(rsq12,rinv12);
1945
1946             /* EWALD ELECTROSTATICS */
1947
1948             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1949             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1950             ewitab           = _mm256_cvttpd_epi32(ewrt);
1951             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1952             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1953                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1954                                             &ewtabF,&ewtabFn);
1955             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1956             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1957
1958             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
1959
1960             fscal            = felec;
1961
1962             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1963
1964             /* Calculate temporary vectorial force */
1965             tx               = _mm256_mul_pd(fscal,dx12);
1966             ty               = _mm256_mul_pd(fscal,dy12);
1967             tz               = _mm256_mul_pd(fscal,dz12);
1968
1969             /* Update vectorial force */
1970             fix1             = _mm256_add_pd(fix1,tx);
1971             fiy1             = _mm256_add_pd(fiy1,ty);
1972             fiz1             = _mm256_add_pd(fiz1,tz);
1973
1974             fjx2             = _mm256_add_pd(fjx2,tx);
1975             fjy2             = _mm256_add_pd(fjy2,ty);
1976             fjz2             = _mm256_add_pd(fjz2,tz);
1977
1978             }
1979
1980             /**************************
1981              * CALCULATE INTERACTIONS *
1982              **************************/
1983
1984             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1985             {
1986
1987             r20              = _mm256_mul_pd(rsq20,rinv20);
1988
1989             /* EWALD ELECTROSTATICS */
1990
1991             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1992             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1993             ewitab           = _mm256_cvttpd_epi32(ewrt);
1994             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1995             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1996                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1997                                             &ewtabF,&ewtabFn);
1998             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1999             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2000
2001             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
2002
2003             fscal            = felec;
2004
2005             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2006
2007             /* Calculate temporary vectorial force */
2008             tx               = _mm256_mul_pd(fscal,dx20);
2009             ty               = _mm256_mul_pd(fscal,dy20);
2010             tz               = _mm256_mul_pd(fscal,dz20);
2011
2012             /* Update vectorial force */
2013             fix2             = _mm256_add_pd(fix2,tx);
2014             fiy2             = _mm256_add_pd(fiy2,ty);
2015             fiz2             = _mm256_add_pd(fiz2,tz);
2016
2017             fjx0             = _mm256_add_pd(fjx0,tx);
2018             fjy0             = _mm256_add_pd(fjy0,ty);
2019             fjz0             = _mm256_add_pd(fjz0,tz);
2020
2021             }
2022
2023             /**************************
2024              * CALCULATE INTERACTIONS *
2025              **************************/
2026
2027             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2028             {
2029
2030             r21              = _mm256_mul_pd(rsq21,rinv21);
2031
2032             /* EWALD ELECTROSTATICS */
2033
2034             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2035             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2036             ewitab           = _mm256_cvttpd_epi32(ewrt);
2037             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2038             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2039                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2040                                             &ewtabF,&ewtabFn);
2041             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2042             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2043
2044             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2045
2046             fscal            = felec;
2047
2048             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2049
2050             /* Calculate temporary vectorial force */
2051             tx               = _mm256_mul_pd(fscal,dx21);
2052             ty               = _mm256_mul_pd(fscal,dy21);
2053             tz               = _mm256_mul_pd(fscal,dz21);
2054
2055             /* Update vectorial force */
2056             fix2             = _mm256_add_pd(fix2,tx);
2057             fiy2             = _mm256_add_pd(fiy2,ty);
2058             fiz2             = _mm256_add_pd(fiz2,tz);
2059
2060             fjx1             = _mm256_add_pd(fjx1,tx);
2061             fjy1             = _mm256_add_pd(fjy1,ty);
2062             fjz1             = _mm256_add_pd(fjz1,tz);
2063
2064             }
2065
2066             /**************************
2067              * CALCULATE INTERACTIONS *
2068              **************************/
2069
2070             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2071             {
2072
2073             r22              = _mm256_mul_pd(rsq22,rinv22);
2074
2075             /* EWALD ELECTROSTATICS */
2076
2077             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2078             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2079             ewitab           = _mm256_cvttpd_epi32(ewrt);
2080             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2081             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2082                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2083                                             &ewtabF,&ewtabFn);
2084             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2085             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2086
2087             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2088
2089             fscal            = felec;
2090
2091             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2092
2093             /* Calculate temporary vectorial force */
2094             tx               = _mm256_mul_pd(fscal,dx22);
2095             ty               = _mm256_mul_pd(fscal,dy22);
2096             tz               = _mm256_mul_pd(fscal,dz22);
2097
2098             /* Update vectorial force */
2099             fix2             = _mm256_add_pd(fix2,tx);
2100             fiy2             = _mm256_add_pd(fiy2,ty);
2101             fiz2             = _mm256_add_pd(fiz2,tz);
2102
2103             fjx2             = _mm256_add_pd(fjx2,tx);
2104             fjy2             = _mm256_add_pd(fjy2,ty);
2105             fjz2             = _mm256_add_pd(fjz2,tz);
2106
2107             }
2108
2109             fjptrA             = f+j_coord_offsetA;
2110             fjptrB             = f+j_coord_offsetB;
2111             fjptrC             = f+j_coord_offsetC;
2112             fjptrD             = f+j_coord_offsetD;
2113
2114             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2115                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2116
2117             /* Inner loop uses 358 flops */
2118         }
2119
2120         if(jidx<j_index_end)
2121         {
2122
2123             /* Get j neighbor index, and coordinate index */
2124             jnrlistA         = jjnr[jidx];
2125             jnrlistB         = jjnr[jidx+1];
2126             jnrlistC         = jjnr[jidx+2];
2127             jnrlistD         = jjnr[jidx+3];
2128             /* Sign of each element will be negative for non-real atoms.
2129              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2130              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
2131              */
2132             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2133
2134             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
2135             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
2136             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
2137
2138             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2139             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2140             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2141             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2142             j_coord_offsetA  = DIM*jnrA;
2143             j_coord_offsetB  = DIM*jnrB;
2144             j_coord_offsetC  = DIM*jnrC;
2145             j_coord_offsetD  = DIM*jnrD;
2146
2147             /* load j atom coordinates */
2148             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
2149                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2150                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2151
2152             /* Calculate displacement vector */
2153             dx00             = _mm256_sub_pd(ix0,jx0);
2154             dy00             = _mm256_sub_pd(iy0,jy0);
2155             dz00             = _mm256_sub_pd(iz0,jz0);
2156             dx01             = _mm256_sub_pd(ix0,jx1);
2157             dy01             = _mm256_sub_pd(iy0,jy1);
2158             dz01             = _mm256_sub_pd(iz0,jz1);
2159             dx02             = _mm256_sub_pd(ix0,jx2);
2160             dy02             = _mm256_sub_pd(iy0,jy2);
2161             dz02             = _mm256_sub_pd(iz0,jz2);
2162             dx10             = _mm256_sub_pd(ix1,jx0);
2163             dy10             = _mm256_sub_pd(iy1,jy0);
2164             dz10             = _mm256_sub_pd(iz1,jz0);
2165             dx11             = _mm256_sub_pd(ix1,jx1);
2166             dy11             = _mm256_sub_pd(iy1,jy1);
2167             dz11             = _mm256_sub_pd(iz1,jz1);
2168             dx12             = _mm256_sub_pd(ix1,jx2);
2169             dy12             = _mm256_sub_pd(iy1,jy2);
2170             dz12             = _mm256_sub_pd(iz1,jz2);
2171             dx20             = _mm256_sub_pd(ix2,jx0);
2172             dy20             = _mm256_sub_pd(iy2,jy0);
2173             dz20             = _mm256_sub_pd(iz2,jz0);
2174             dx21             = _mm256_sub_pd(ix2,jx1);
2175             dy21             = _mm256_sub_pd(iy2,jy1);
2176             dz21             = _mm256_sub_pd(iz2,jz1);
2177             dx22             = _mm256_sub_pd(ix2,jx2);
2178             dy22             = _mm256_sub_pd(iy2,jy2);
2179             dz22             = _mm256_sub_pd(iz2,jz2);
2180
2181             /* Calculate squared distance and things based on it */
2182             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2183             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
2184             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
2185             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
2186             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2187             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2188             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
2189             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2190             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2191
2192             rinv00           = avx256_invsqrt_d(rsq00);
2193             rinv01           = avx256_invsqrt_d(rsq01);
2194             rinv02           = avx256_invsqrt_d(rsq02);
2195             rinv10           = avx256_invsqrt_d(rsq10);
2196             rinv11           = avx256_invsqrt_d(rsq11);
2197             rinv12           = avx256_invsqrt_d(rsq12);
2198             rinv20           = avx256_invsqrt_d(rsq20);
2199             rinv21           = avx256_invsqrt_d(rsq21);
2200             rinv22           = avx256_invsqrt_d(rsq22);
2201
2202             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
2203             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
2204             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
2205             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
2206             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
2207             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
2208             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
2209             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
2210             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
2211
2212             fjx0             = _mm256_setzero_pd();
2213             fjy0             = _mm256_setzero_pd();
2214             fjz0             = _mm256_setzero_pd();
2215             fjx1             = _mm256_setzero_pd();
2216             fjy1             = _mm256_setzero_pd();
2217             fjz1             = _mm256_setzero_pd();
2218             fjx2             = _mm256_setzero_pd();
2219             fjy2             = _mm256_setzero_pd();
2220             fjz2             = _mm256_setzero_pd();
2221
2222             /**************************
2223              * CALCULATE INTERACTIONS *
2224              **************************/
2225
2226             if (gmx_mm256_any_lt(rsq00,rcutoff2))
2227             {
2228
2229             r00              = _mm256_mul_pd(rsq00,rinv00);
2230             r00              = _mm256_andnot_pd(dummy_mask,r00);
2231
2232             /* EWALD ELECTROSTATICS */
2233
2234             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2235             ewrt             = _mm256_mul_pd(r00,ewtabscale);
2236             ewitab           = _mm256_cvttpd_epi32(ewrt);
2237             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2238             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2239                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2240                                             &ewtabF,&ewtabFn);
2241             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2242             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
2243
2244             /* LENNARD-JONES DISPERSION/REPULSION */
2245
2246             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2247             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2248
2249             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
2250
2251             fscal            = _mm256_add_pd(felec,fvdw);
2252
2253             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2254
2255             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2256
2257             /* Calculate temporary vectorial force */
2258             tx               = _mm256_mul_pd(fscal,dx00);
2259             ty               = _mm256_mul_pd(fscal,dy00);
2260             tz               = _mm256_mul_pd(fscal,dz00);
2261
2262             /* Update vectorial force */
2263             fix0             = _mm256_add_pd(fix0,tx);
2264             fiy0             = _mm256_add_pd(fiy0,ty);
2265             fiz0             = _mm256_add_pd(fiz0,tz);
2266
2267             fjx0             = _mm256_add_pd(fjx0,tx);
2268             fjy0             = _mm256_add_pd(fjy0,ty);
2269             fjz0             = _mm256_add_pd(fjz0,tz);
2270
2271             }
2272
2273             /**************************
2274              * CALCULATE INTERACTIONS *
2275              **************************/
2276
2277             if (gmx_mm256_any_lt(rsq01,rcutoff2))
2278             {
2279
2280             r01              = _mm256_mul_pd(rsq01,rinv01);
2281             r01              = _mm256_andnot_pd(dummy_mask,r01);
2282
2283             /* EWALD ELECTROSTATICS */
2284
2285             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2286             ewrt             = _mm256_mul_pd(r01,ewtabscale);
2287             ewitab           = _mm256_cvttpd_epi32(ewrt);
2288             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2289             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2290                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2291                                             &ewtabF,&ewtabFn);
2292             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2293             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
2294
2295             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
2296
2297             fscal            = felec;
2298
2299             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2300
2301             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2302
2303             /* Calculate temporary vectorial force */
2304             tx               = _mm256_mul_pd(fscal,dx01);
2305             ty               = _mm256_mul_pd(fscal,dy01);
2306             tz               = _mm256_mul_pd(fscal,dz01);
2307
2308             /* Update vectorial force */
2309             fix0             = _mm256_add_pd(fix0,tx);
2310             fiy0             = _mm256_add_pd(fiy0,ty);
2311             fiz0             = _mm256_add_pd(fiz0,tz);
2312
2313             fjx1             = _mm256_add_pd(fjx1,tx);
2314             fjy1             = _mm256_add_pd(fjy1,ty);
2315             fjz1             = _mm256_add_pd(fjz1,tz);
2316
2317             }
2318
2319             /**************************
2320              * CALCULATE INTERACTIONS *
2321              **************************/
2322
2323             if (gmx_mm256_any_lt(rsq02,rcutoff2))
2324             {
2325
2326             r02              = _mm256_mul_pd(rsq02,rinv02);
2327             r02              = _mm256_andnot_pd(dummy_mask,r02);
2328
2329             /* EWALD ELECTROSTATICS */
2330
2331             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2332             ewrt             = _mm256_mul_pd(r02,ewtabscale);
2333             ewitab           = _mm256_cvttpd_epi32(ewrt);
2334             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2335             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2336                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2337                                             &ewtabF,&ewtabFn);
2338             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2339             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
2340
2341             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
2342
2343             fscal            = felec;
2344
2345             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2346
2347             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2348
2349             /* Calculate temporary vectorial force */
2350             tx               = _mm256_mul_pd(fscal,dx02);
2351             ty               = _mm256_mul_pd(fscal,dy02);
2352             tz               = _mm256_mul_pd(fscal,dz02);
2353
2354             /* Update vectorial force */
2355             fix0             = _mm256_add_pd(fix0,tx);
2356             fiy0             = _mm256_add_pd(fiy0,ty);
2357             fiz0             = _mm256_add_pd(fiz0,tz);
2358
2359             fjx2             = _mm256_add_pd(fjx2,tx);
2360             fjy2             = _mm256_add_pd(fjy2,ty);
2361             fjz2             = _mm256_add_pd(fjz2,tz);
2362
2363             }
2364
2365             /**************************
2366              * CALCULATE INTERACTIONS *
2367              **************************/
2368
2369             if (gmx_mm256_any_lt(rsq10,rcutoff2))
2370             {
2371
2372             r10              = _mm256_mul_pd(rsq10,rinv10);
2373             r10              = _mm256_andnot_pd(dummy_mask,r10);
2374
2375             /* EWALD ELECTROSTATICS */
2376
2377             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2378             ewrt             = _mm256_mul_pd(r10,ewtabscale);
2379             ewitab           = _mm256_cvttpd_epi32(ewrt);
2380             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2381             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2382                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2383                                             &ewtabF,&ewtabFn);
2384             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2385             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
2386
2387             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
2388
2389             fscal            = felec;
2390
2391             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2392
2393             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2394
2395             /* Calculate temporary vectorial force */
2396             tx               = _mm256_mul_pd(fscal,dx10);
2397             ty               = _mm256_mul_pd(fscal,dy10);
2398             tz               = _mm256_mul_pd(fscal,dz10);
2399
2400             /* Update vectorial force */
2401             fix1             = _mm256_add_pd(fix1,tx);
2402             fiy1             = _mm256_add_pd(fiy1,ty);
2403             fiz1             = _mm256_add_pd(fiz1,tz);
2404
2405             fjx0             = _mm256_add_pd(fjx0,tx);
2406             fjy0             = _mm256_add_pd(fjy0,ty);
2407             fjz0             = _mm256_add_pd(fjz0,tz);
2408
2409             }
2410
2411             /**************************
2412              * CALCULATE INTERACTIONS *
2413              **************************/
2414
2415             if (gmx_mm256_any_lt(rsq11,rcutoff2))
2416             {
2417
2418             r11              = _mm256_mul_pd(rsq11,rinv11);
2419             r11              = _mm256_andnot_pd(dummy_mask,r11);
2420
2421             /* EWALD ELECTROSTATICS */
2422
2423             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2424             ewrt             = _mm256_mul_pd(r11,ewtabscale);
2425             ewitab           = _mm256_cvttpd_epi32(ewrt);
2426             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2427             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2428                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2429                                             &ewtabF,&ewtabFn);
2430             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2431             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2432
2433             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
2434
2435             fscal            = felec;
2436
2437             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2438
2439             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2440
2441             /* Calculate temporary vectorial force */
2442             tx               = _mm256_mul_pd(fscal,dx11);
2443             ty               = _mm256_mul_pd(fscal,dy11);
2444             tz               = _mm256_mul_pd(fscal,dz11);
2445
2446             /* Update vectorial force */
2447             fix1             = _mm256_add_pd(fix1,tx);
2448             fiy1             = _mm256_add_pd(fiy1,ty);
2449             fiz1             = _mm256_add_pd(fiz1,tz);
2450
2451             fjx1             = _mm256_add_pd(fjx1,tx);
2452             fjy1             = _mm256_add_pd(fjy1,ty);
2453             fjz1             = _mm256_add_pd(fjz1,tz);
2454
2455             }
2456
2457             /**************************
2458              * CALCULATE INTERACTIONS *
2459              **************************/
2460
2461             if (gmx_mm256_any_lt(rsq12,rcutoff2))
2462             {
2463
2464             r12              = _mm256_mul_pd(rsq12,rinv12);
2465             r12              = _mm256_andnot_pd(dummy_mask,r12);
2466
2467             /* EWALD ELECTROSTATICS */
2468
2469             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2470             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2471             ewitab           = _mm256_cvttpd_epi32(ewrt);
2472             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2473             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2474                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2475                                             &ewtabF,&ewtabFn);
2476             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2477             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2478
2479             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
2480
2481             fscal            = felec;
2482
2483             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2484
2485             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2486
2487             /* Calculate temporary vectorial force */
2488             tx               = _mm256_mul_pd(fscal,dx12);
2489             ty               = _mm256_mul_pd(fscal,dy12);
2490             tz               = _mm256_mul_pd(fscal,dz12);
2491
2492             /* Update vectorial force */
2493             fix1             = _mm256_add_pd(fix1,tx);
2494             fiy1             = _mm256_add_pd(fiy1,ty);
2495             fiz1             = _mm256_add_pd(fiz1,tz);
2496
2497             fjx2             = _mm256_add_pd(fjx2,tx);
2498             fjy2             = _mm256_add_pd(fjy2,ty);
2499             fjz2             = _mm256_add_pd(fjz2,tz);
2500
2501             }
2502
2503             /**************************
2504              * CALCULATE INTERACTIONS *
2505              **************************/
2506
2507             if (gmx_mm256_any_lt(rsq20,rcutoff2))
2508             {
2509
2510             r20              = _mm256_mul_pd(rsq20,rinv20);
2511             r20              = _mm256_andnot_pd(dummy_mask,r20);
2512
2513             /* EWALD ELECTROSTATICS */
2514
2515             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2516             ewrt             = _mm256_mul_pd(r20,ewtabscale);
2517             ewitab           = _mm256_cvttpd_epi32(ewrt);
2518             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2519             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2520                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2521                                             &ewtabF,&ewtabFn);
2522             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2523             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2524
2525             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
2526
2527             fscal            = felec;
2528
2529             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2530
2531             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2532
2533             /* Calculate temporary vectorial force */
2534             tx               = _mm256_mul_pd(fscal,dx20);
2535             ty               = _mm256_mul_pd(fscal,dy20);
2536             tz               = _mm256_mul_pd(fscal,dz20);
2537
2538             /* Update vectorial force */
2539             fix2             = _mm256_add_pd(fix2,tx);
2540             fiy2             = _mm256_add_pd(fiy2,ty);
2541             fiz2             = _mm256_add_pd(fiz2,tz);
2542
2543             fjx0             = _mm256_add_pd(fjx0,tx);
2544             fjy0             = _mm256_add_pd(fjy0,ty);
2545             fjz0             = _mm256_add_pd(fjz0,tz);
2546
2547             }
2548
2549             /**************************
2550              * CALCULATE INTERACTIONS *
2551              **************************/
2552
2553             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2554             {
2555
2556             r21              = _mm256_mul_pd(rsq21,rinv21);
2557             r21              = _mm256_andnot_pd(dummy_mask,r21);
2558
2559             /* EWALD ELECTROSTATICS */
2560
2561             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2562             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2563             ewitab           = _mm256_cvttpd_epi32(ewrt);
2564             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2565             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2566                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2567                                             &ewtabF,&ewtabFn);
2568             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2569             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2570
2571             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2572
2573             fscal            = felec;
2574
2575             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2576
2577             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2578
2579             /* Calculate temporary vectorial force */
2580             tx               = _mm256_mul_pd(fscal,dx21);
2581             ty               = _mm256_mul_pd(fscal,dy21);
2582             tz               = _mm256_mul_pd(fscal,dz21);
2583
2584             /* Update vectorial force */
2585             fix2             = _mm256_add_pd(fix2,tx);
2586             fiy2             = _mm256_add_pd(fiy2,ty);
2587             fiz2             = _mm256_add_pd(fiz2,tz);
2588
2589             fjx1             = _mm256_add_pd(fjx1,tx);
2590             fjy1             = _mm256_add_pd(fjy1,ty);
2591             fjz1             = _mm256_add_pd(fjz1,tz);
2592
2593             }
2594
2595             /**************************
2596              * CALCULATE INTERACTIONS *
2597              **************************/
2598
2599             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2600             {
2601
2602             r22              = _mm256_mul_pd(rsq22,rinv22);
2603             r22              = _mm256_andnot_pd(dummy_mask,r22);
2604
2605             /* EWALD ELECTROSTATICS */
2606
2607             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2608             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2609             ewitab           = _mm256_cvttpd_epi32(ewrt);
2610             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2611             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2612                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2613                                             &ewtabF,&ewtabFn);
2614             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2615             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2616
2617             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2618
2619             fscal            = felec;
2620
2621             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2622
2623             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2624
2625             /* Calculate temporary vectorial force */
2626             tx               = _mm256_mul_pd(fscal,dx22);
2627             ty               = _mm256_mul_pd(fscal,dy22);
2628             tz               = _mm256_mul_pd(fscal,dz22);
2629
2630             /* Update vectorial force */
2631             fix2             = _mm256_add_pd(fix2,tx);
2632             fiy2             = _mm256_add_pd(fiy2,ty);
2633             fiz2             = _mm256_add_pd(fiz2,tz);
2634
2635             fjx2             = _mm256_add_pd(fjx2,tx);
2636             fjy2             = _mm256_add_pd(fjy2,ty);
2637             fjz2             = _mm256_add_pd(fjz2,tz);
2638
2639             }
2640
2641             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2642             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2643             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2644             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2645
2646             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2647                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2648
2649             /* Inner loop uses 367 flops */
2650         }
2651
2652         /* End of innermost loop */
2653
2654         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2655                                                  f+i_coord_offset,fshift+i_shift_offset);
2656
2657         /* Increment number of inner iterations */
2658         inneriter                  += j_index_end - j_index_start;
2659
2660         /* Outer loop uses 18 flops */
2661     }
2662
2663     /* Increment number of outer iterations */
2664     outeriter        += nri;
2665
2666     /* Update outer/inner flops */
2667
2668     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*367);
2669 }